Recognition of the dangerous driving behaviors and the driving styles in weaving areas based on a hybrid neural network

doi:10.3969/j.issn.1674-8484.2025.05.003

Abstract

Abstract:

A hybrid neural network analysis method was proposed based on historical trajectory data to identify and predict dangerous driving behaviors and driving styles in weaving areas. The different dangerous driving behaviors were clustered and analyzed by using the K-means++ algorithm with the key features of the longitudinal velocity, the lateral acceleration, and the longitudinal acceleration being extracted to characterize different driving styles. A driving style prediction model was constructed based on a Long Short-Term Memory (LSTM) network, a Convolutional Neural Network (CNN), and a Knowledge-Attention Network (KAN), with conducting digital simulations and ablation comparison experiments. The results show that the model has the Area Under Curve (AUC) of the Receiver Operating Characteristic (ROC), a dimension-one quantity, of 0.846. the model's classification and prediction accuracy of dangerous driving behaviors and driving styles increased by 6.73%, 3.12%, and 4.72%, while increasing the generalization verification accuracy by 6.3%, 2.5%, and 3.9%, compared with models using LSTM, CNN-LSTM, and LSTM-KAN.

Key words: Intelligent transport, urban expressway weaving area, dangerous driving behaviors, driving styles, k-means++ algorithm, cluster analysis, deep learning, hybrid neural

CLC Number:

U491

CHENG Zeyang, DUAN Yiyang, YANG Mengmeng, FENG Zhongxiang, WANG He, ZHU Xiaojun, BAO Lixia. Recognition of the dangerous driving behaviors and the driving styles in weaving areas based on a hybrid neural network[J]. Journal of Automotive Safety and Energy, 2025, 16(5): 688-697.

Figures/Tables 17

References 17

[1]	张巍. 基于模型预测和路径规划的汽车主动转向避撞控制研究[D]. 重庆: 重庆大学, 2017.
	ZHANG Wei. Research on vehicle active steering collision avoidance control based on model prediction and path planning[D]. Chongqing: Chongqing University, 2017. (in Chinese)
[2]	LIN Qiufeng, Ulsoy A G. Vehicle dynamics and external disturbance estimation for vehicle path prediction[J]. IEEE Transport Contr Syst Tech, 2000, 8(3): 508-518.
[3]	Polychronopoulos A, Tsogas M, Amditis A J, et al. Sensor fusion for predicting vehicles' path for collision avoidance systems[J]. IEEE Trans Intell Transport Syst, 2007, 8(3): 549-562. doi: 10.1109/TITS.2007.903439 URL
[4]	LI Jian, MA Hao, ZHAN Wei, et al. Generic Probabilistic Interactive Situation Recognition and Prediction:From Virtual to Real[C]//2018 21st Int’l Conf Intell Transp Syst (ITSC), IEEE ITSC, 2018: 2-3.
[5]	谷新平, 韩云鹏, 于俊甫. 基于决策机理与支持向量机的车辆换道决策模型[J]. 哈尔滨工业大学学报, 2020, 52(7): 117-127.
	GU Xinping, HAN Yunpeng, YU Junfu. Vehicle lane-changing decision model based on decision mechanism and support vector machine[J]. J Harbin Inst Techn, 2020, 52(7): 117-127. (in Chinese)
[6]	刘志强, 吴雪刚, 倪捷. 基于HMM和SVM级联算法的驾驶意图识别[J]. 汽车工程, 2018, 40(7): 858-864.
	LIU Zhiqiang, WU Xuegang, NI Jie. Driving intention recognition based on HMM and SVM cascade algorithm[J]. Automo Engineering, 2018, 40(7): 858-864. (in Chinese)
[7]	王运运, 尹慧琳. 基于 AIOHMM 模型的驾驶行为预测[J]. 信息通信, 2019, 3: 120-123.
	WANG Yunyun, YIN Huilin. Driving behavior prediction based on AIOHMM model[J]. Info Commu, 2019, 3: 120-123. (in Chinese)
[8]	DENG Qiang, SÖFFKER D. Improved driving behaviors prediction based on fuzzy logic-hidden Markov model (FL-HMM)[C]//2018 IEEE Intell Vehi Symp (IV), 2018: 2003-2008.
[9]	王炜炜, 施笑畏. 大交通流的红绿灯十字路口驾驶行为预测研究[J]. 物流科技, 2018, 41(4): 71-76.
	WANG Weiwei, SHI Xiaowei. Research on prediction of driving behavior at traffic light intersections with large traffic flow[J]. Logis Sci Tech, 2018, 41(4): 71-76. (in Chinese)
[10]	HU Yihan, ZHAN Wei, TOMIZUKA M. A framework for probabilistic generic traffic scene prediction[C]//21st Int’l Conf Intell Transp Syst (ITSC), IEEE ITSC 2018, 154: 2790-2796.
[11]	季学武, 费聪, 何祥坤, 等. 基于 LSTM 网络的驾驶意图识别及车辆轨迹预测[J]. 中国公路学报, 2019, 32(6): 34-42. doi: 10.19721/j.cnki.1001-7372.2019.06.003
	JI Xuewu, FEI Cong, HE Xiangkun, et al. Driving intention recognition and vehicle trajectory prediction based on LSTM network[J]. Chin J Highw, 2019, 32(6): 34-42. (in Chinese)
[12]	张海民. 基于深度学习模型的疲劳驾驶行为识别算法[J]. 汽车安全与节能学报, 2024, 15(1): 121-128.
	ZHANG Haimin. Fatigue driving behavior recognition algorithm based on deep learning model[J]. J Automo Safe Energ, 2024, 15(1): 121-128. (in Chinese)
[13]	Diehl F, Brunner T, Le MT, et al. Graph neural networks for modelling traffic participant interaction[C]//2019 IEEE Intell Vehi Symp (IV). IEEE, 2019: 695-701.
[14]	LI Xin, YING Xiaowen, CHUAH Mooi Choo. GRIP:Graph-based Interaction-aware trajectory prediction[C]//2019 IEEE Intell Transp Syst Conf (ITSC). IEEE, 2019: 3960-3966.
[15]	张志鸿. 基于 OBD 数据分析的驾驶行为研究[D]. 西安: 长安大学, 2017.
	ZHANG Zhihong. Research on driving behavior based on OBD data analysis[D]. Xi'an: Chang'an University, 2017. (in Chinese)
[16]	Arthur D, Vassilvitskii S. K-Means++: The advantages of careful seeding[C]//Proc 18^th Annu ACM-SIAM Symp Discrete Algorithms (SODA). New Orleans, USA, 2007: 1027-1035.
[17]	何荣, 郭睿, 管欣, 等. 钢板弹簧 Fancher 模型物理机理及参数辨识[J]. 吉林大学学报: 工学版, 2013, 43(1): 12-16.
	HE Rong, GUO Rui, GUAN Xin, et al. Physical mechanism and parameter identification of leaf spring Fancher model[J]. J Jilin Univ: Eng Ed, 2013, 43(1): 12-16. (in Chinese)

类型	特征	特征参数
超速	车辆速度超过限制速度	车速(v_y)、限速 (v_th)
急加速	短时间内纵向加速度偏大	纵向加速度(a_y)
急减速	短时间内纵向减速度偏大	纵向加速度(a_y)
急变道	短时间内车辆快速变道过快	横向速度(v_x)

类型	特征	特征参数
超速	车辆速度超过限制速度	车速(v_y)、限速 (v_th)
急加速	短时间内纵向加速度偏大	纵向加速度(a_y)
急减速	短时间内纵向减速度偏大	纵向加速度(a_y)
急变道	短时间内车辆快速变道过快	横向速度(v_x)

驾驶风格	特征
危险型	存在任意危险驾驶行为(如超速、急变道、急加减速)。
风险型	不存在危险驾驶行为，存在任意危险驾驶行为趋势。
安全型	不存在危险驾驶行为，也不存在危险驾驶行为趋势。

驾驶风格	特征
危险型	存在任意危险驾驶行为(如超速、急变道、急加减速)。
风险型	不存在危险驾驶行为，存在任意危险驾驶行为趋势。
安全型	不存在危险驾驶行为，也不存在危险驾驶行为趋势。

模型	指标
模型	Accuracy	Recall	F₁	AUC
LSTM	0.798	0.803	0.801	0.803
CNN-LSTM	0.822	0.826	0.824	0.826
LSTM-KAN	0.813	0.817	0.815	0.817
LSTM-CNN-KAN	0.843	0.846	0.844	0.846